MEMS technology has been under steady development for some time, and as a result various MEMS devices have been considered and demonstrated for several applications. MEMS technology is an attractive approach for providing inertial sensors, such as accelerometers for measuring linear acceleration and gyroscopes for measuring angular velocity. A MEMS inertial sensor typically includes a proof mass which is flexibly attached to the rest of the device. Relative motion between the proof mass and the rest of the device is driven by actuators and/or sensed by sensors in various ways, depending on the detailed device design. Other MEMS applications include optical applications such as movable mirrors, and RF applications such as RF switches and resonators.
Since MEMS fabrication technology is typically based on processing planar silicon wafers, it is useful to classify MEMS devices according to whether the actuation and/or sensing performed in an inertial sensor (or other application) is in-plane or out of plane (Le., vertical). More specifically, a device is “in-plane” if all of its sensing and/or actuation is in-plane, otherwise it is “vertical”. Thus MEMS devices are undergoing steady development, despite fabrication difficulties that tend to increase.
One approach which has been used to fabricate vertical MEMS devices is hybrid integration, where elements of a MEMS assembly are individually assembled to form the desired vertical structure. For example, attachment of a spacer to a substrate, followed by attachment of a deformable diaphragm to the spacer, provides a vertical MEMS structure having a spacing between diaphragm and substrate controlled by the spacer. U.S. Pat. No. 6,426,687 provides further information on this approach. Although hybrid integration can provide vertical MEMS devices, the cost tends to be high, since manual processing steps are usually required, and because hybrid integration is typically performed on single devices. Therefore, there is a need for reduced cost integrated MEMS devices that is unmet in the prior art.
CMOS compatible wafer-wafer bonding is very desirable for wafer-level-packaging. Its use has been demonstrated in a variety of different technologies. However, most of these processes have been limited to providing protection of a sensitive feature from post process handling, such as sawing, die bonding, testing, package, etc.
The need for a robust wafer level integration that can allow for simultaneous wafer-level-packaging and electrical interconnect is very high and can open up a multitude of new smaller, low-cost and feature rich MEMS products. The following describes conventional methods for bonding and their problems.
Organic or Adhesive Based Methods
Materials such as Benzocyclobutene (BCB), polyamide, photo resists, patternable RTV, and others have been spun on and used to form permanent bonds between wafers. These materials have disadvantages in that because they are organic, they tend to outgas and so are unsuitable for forming hermetic enclosures, and also they are susceptible to solvents, or moistures which can lead to problems with long term reliability and drift of a device's performance. Additionally, they are insulating materials and so are incapable of forming a conductive path between two substrates.
One popular method of making wafer-wafer bonding is by use of frit glass. Frit glass is typically screen printed on the cover wafers and reflowed to form a patterned glass interface for subsequent wafer-wafer bonding. Frit glass has a typical melting point near 500° C. and can be remelted post wafer-wafer aligned bond in a special temperature chamber with a controlled environment. The primary use of glass frit is to provide for the cover substrate and a hermetic sealed cavity for the MEMS. Frit glass technology has been utilized in the MEMS industry for many decades. Several major drawbacks are that frit glass does not provide for electrical interconnection between the MEMS and cover, to achieve a hermetic seal interface, minimum of 400 micron seal ring width is required which makes small MEMS devices, such as resonators and optical devices, much larger than otherwise. Also, frit glass screen printed is inherently a thick film process with tens of microns in thickness and several microns of nonuniformity.
Metal-Metal Bonding
CMOS compatible eutectic bonding has been demonstrated with indium-gold, solder-gold, gold-gold, etc. In order to bond a CMOS wafer, all of these prior art systems require the addition of non-standard layers, such as plating of lead, indium, gold, etc., to be added to the CMOS wafer. Although these processes are capable of hermetic seals and electrical interface, achieving fine features, small gaps and wafer uniformity is very challenging and will result in yield losses.
There are many MEMS device applications that require an electro-mechanical interface between the CMOS substrate and the MEMS substrates that are in micron gaps and require submicron uniformity. Most plating processes require under-layer barrier metalization with tens of microns thickness, and uniformity across the wafer is measured in microns. Hence it is not possible to specify one or two micron gap controls between the MEMS and CMOS substrates using this bonding methodology.
The ability to make high density and reliable electrical contacts between the MEMS and CMOS substrates can be very beneficial and provide for an all new generation of MEMS devices with much added functionality, smart electronics, smaller size, and lower cost. Finally, it is important to provide a lead free alloy based upon environmental considerations.
Accordingly, what is needed is a system and method for providing wafer bonding that overcomes the above-identified problems. The system and method should be easily implemented, cost effective and adaptable to existing bonding processes. The present invention addresses such a need.